Materials of Micro- and Nanotechnologies

Atomic bonding and crystal structure

Ionic bonding transfers charge, covalent bonding shares electrons, and metallic bonding permits delocalized electrons. Real solids can combine these ideal descriptions. Bonding shapes stiffness, melting behaviour and electronic states. A crystal repeats a lattice plus a basis; the unit cell is a bookkeeping choice, not a separate physical object.

Miller indices describe plane orientation using reciprocals of axis intercepts. An infinite intercept contributes zero. Directions use square brackets and planes use parentheses; in non-cubic crystals a direction and the identically indexed plane normal need not coincide.

Build the physical picture

Begin with the competition between attraction and repulsion. Atoms approach until the total potential energy reaches a minimum; moving either closer or farther costs energy. The curvature near that minimum controls elastic stiffness. This connects atomic bonding to a property measured on a macroscopic specimen.

A lattice specifies translations; a basis specifies which atoms sit at each lattice point. Silicon has a diamond-cubic structure, not a simple cubic arrangement of one atom per corner. Using a cubic unit cell does not imply simple-cubic bonding.

From mechanism to a usable model

Separate three geometric questions: nearest-neighbour distance, spacing between parallel planes, and repeat distance along a direction. They are generally different. Diffraction measures reciprocal-space information associated with periodic planes, while an atomic image may resolve projected columns.

For Miller indices, choose intercepts in units of the lattice parameters, invert them, and multiply to the smallest integer set. A negative intercept gives a negative index. Translating a plane to avoid an origin intercept changes its position but not its orientation.

d_hkl=a/√(h²+k²+l²)

Model assumptions: Cubic lattice only; a and d use the same length unit.

Construct a plane before calculating its spacing

Sketch the axes and mark intercepts in lattice units.

Invert finite intercepts; use zero for an infinite intercept. Clear fractions.

For a cubic crystal only, insert the resulting integers into d=a/√(h²+k²+l²).

Mistake to avoid: A unit-cell drawing is a representation of an extended solid; its edges are not physical grain boundaries.

Worked problem: For a cubic lattice with a=0.543 nm, find d₁₁₁.

d=0.543/√3=0.3135 nm. This is a plane spacing, not a nearest-neighbour distance.

Self-test: Find indices for a plane intercepting axes at a, 2a and infinity.

Reciprocals are 1, 1/2, 0. Clear fractions to obtain (210).

A cubic material has a=0.4 nm. Compare d100 and d200. Do they describe the same orientation?

d100=0.4 nm and d200=0.2 nm. Both normals point along the same axis, but the indexed spacings differ. Orientation families and diffraction orders must not be confused.

Properties of engineering materials

Material properties arise from structure and respond to temperature, frequency, stress and composition. Electrical conductivity describes charge transport; thermal conductivity describes heat transport. Young’s modulus describes small elastic strain, while strength describes a failure or yielding condition. They are different measurements.

Optical properties depend on allowed excitations and wavelength. A material transparent in one spectral range may absorb strongly in another. Select properties under actual operating conditions and distinguish bulk values from thin-film values changed by grain size, interfaces or porosity.

Build the physical picture

A property is the coefficient connecting an applied cause to a measured response. Electrical resistivity connects current density and electric field; thermal conductivity connects heat flux and temperature gradient. Even if electrons contribute to both responses, the two coefficients have different units and are not interchangeable.

Thin films contain surfaces, grain boundaries and sometimes porosity. These alter scattering, stiffness and heat flow. A handbook value for a dense bulk crystal is a starting estimate, not a guaranteed film measurement.

From mechanism to a usable model

First identify the geometry. A film resistor of length L, width w and thickness t has R=ρL/(wt). Defining sheet resistance R□=ρ/t separates the material and thickness from the number of squares L/w. Contact resistance must be added separately.

For mechanics, the linear slope E applies only before significant plasticity, cracking or nonlinear deformation. A stiff material can still be brittle. For optical properties, specify wavelength and distinguish absorption from reflection and scattering.

σ=1/ρ; stress=E·strain; heat flux=−k_th∇T

Model assumptions: Ohmic, linear-elastic and Fourier heat-flow regimes; each has its own validity limits.

Extract a material coefficient from a measurement

Measure dimensions and state temperature, frequency and loading conditions.

Remove known contact or instrument contributions before attributing the response to the material.

Rearrange the governing relation and check the coefficient’s units and plausible range.

Mistake to avoid: Low resistance of a wide sample does not establish low resistivity; geometry can explain the result.

Worked problem: A film has resistivity 2×10⁻⁶ Ωm and thickness 100 nm. Find sheet resistance.

R□=ρ/t=2×10⁻⁶/10⁻⁷=20 Ω/square. The film may differ from bulk material because microstructure changes scattering.

Self-test: Why is modulus not the same as fracture strength?

Modulus is the slope of elastic stress versus strain; fracture depends on flaws, geometry and failure mechanisms.

A 20 Ω/square film is patterned into a 200 µm long, 20 µm wide track. What is its ideal resistance?

There are L/w=10 squares, so R=200 Ω. This neglects contacts, corner effects and thickness nonuniformity.

Defects and microstructure

A perfect lattice is a reference model. Vacancies remove atoms, interstitials add atoms between normal sites, and substitutional impurities replace host atoms. Dislocations are line defects that allow plastic deformation; grain boundaries separate differently oriented crystals.

Defects may scatter carriers, enable diffusion, pin domain walls or create electronic trap states. Their effect can be useful or harmful: doping intentionally changes carrier populations, while uncontrolled contamination may shorten lifetime. Describe the defect’s location and physical mechanism before claiming an effect.

Build the physical picture

A vacancy enables an atom to jump into an adjacent empty lattice site. The vacancy effectively moves in the opposite direction. Formation of the vacancy and migration of an atom have separate energy costs; a measured diffusion activation energy may include both.

Dislocations let slip advance locally rather than requiring a whole atomic plane to move at once. Grain boundaries are extended regions of imperfect order. Neither should be described as simply a missing atom.

From mechanism to a usable model

Equilibrium defect concentrations minimize free energy, balancing an energy penalty against configurational entropy. The simple exponential vacancy model omits an entropy prefactor. Cooling quickly can retain a population different from the equilibrium population at the final temperature.

Electronic effects depend on defect energy levels and capture rates. A dopant providing shallow carriers is different from a deep trap that promotes recombination. Identify the defect and its location before drawing a conclusion about lifetime or conductivity.

Equilibrium vacancy fraction≈exp(−E_f/kT)

Model assumptions: Simplified dilute-defect model with entropy prefactor omitted.

Reason from a defect to a device symptom

Classify the defect as point, line, interface or volume.

Identify its coupling to carriers, stress or atomic transport.

Predict a measurable consequence and choose a test that can separate it from competing mechanisms.

Mistake to avoid: All impurities are not beneficial dopants, and a lower defect count does not automatically improve every functional material.

Worked problem: For E_f=1 eV at 1000 K, estimate vacancy fraction.

kT≈0.08617 eV; exp(−1/0.08617)≈9.1×10⁻⁶. This omits formation entropy and interaction effects.

Self-test: Why can grain boundaries accelerate diffusion?

Their disturbed structure can provide lower-barrier pathways than the ordered bulk lattice.

If formation energy is unchanged, does equilibrium vacancy concentration increase or decrease on heating?

It increases: −Ef/(kT) becomes less negative. This is an equilibrium statement; a real sample may need time to exchange atoms and equilibrate.

Thin-film growth

Deposited atoms first adsorb, migrate and form nuclei. Their competition between binding to the substrate and binding to each other produces layer growth, island growth or an initial layer followed by islands. Surface diffusion and deposition rate determine how far atoms can travel before incorporation.

Coalescing islands can leave voids or generate stress. Film properties therefore depend on growth history, substrate preparation and temperature, not just chemical identity. A nominal deposited thickness does not guarantee a continuous conductive film.

Build the physical picture

An arriving atom can remain on the surface, diffuse, desorb or join a cluster. A stable nucleus survives long enough to collect more atoms. The density of such nuclei determines how islands meet and what grain structure remains after coalescence.

Surface and interface energies determine whether spreading a film is favourable. Kinetics determines whether atoms can reach the favourable arrangement. A thermodynamically preferred smooth layer can still grow rough if atoms are immobilized too quickly.

From mechanism to a usable model

Compare surface migration time with the time between arrivals. Increasing temperature can improve migration but also increase desorption or cause reactions. Increasing flux usually reduces the time available for rearrangement. Therefore “hotter produces better films” is not a universal rule.

Electrical continuity requires connected pathways. Before coalescence, nominal thickness represents deposited material per area, not the thickness of a uniform continuous slab. After coalescence, grain boundaries and residual stress still carry information about growth.

Diffusion length≈√(D_s τ)

Model assumptions: Surface random-walk scale; numerical prefactor depends on dimensionality and definition.

Predict a growth change without overclaiming

List the changed variable: temperature, flux or substrate preparation.

Identify which rates and surface energies it can affect.

Predict morphology, then verify coverage, roughness and continuity rather than relying on thickness alone.

Mistake to avoid: The same deposition recipe can produce different morphology on different substrates.

Worked problem: If surface diffusivity increases by a factor of 9 while residence time is unchanged, how does diffusion length change?

It increases by √9=3. Longer migration can alter nucleation density, but morphology also depends on surface energies and incoming flux.

Self-test: Why might a 3 nm metal film conduct poorly?

It may consist of disconnected islands below a percolation threshold rather than a uniform sheet.

Why might resistance fall abruptly while deposited thickness increases smoothly?

Initially isolated islands become connected near a percolation transition. The conductive network changes sharply even though deposited mass rises continuously.

Thin-film deposition

PVD transports material physically from a source: evaporation releases atoms; sputtering transfers momentum from energetic ions. CVD uses chemical reactions of gas precursors at a surface, and PECVD adds plasma activation. ALD separates self-limiting surface reactions with purges.

Select a process using conformality, thermal budget, contamination, stress and throughput. ALD’s cycle-by-cycle control is useful on complex geometry, but requires suitable chemistry and sufficient precursor access. A single growth-per-cycle number does not prove nucleation-free behaviour on every substrate.

Build the physical picture

Evaporation delivers mainly directional flux; sputtering ejects target atoms through ion momentum transfer. A substrate sees a distribution of trajectories, which matters for steps and trenches. CVD instead relies on precursor transport and surface reaction, while ALD separates reactions to make growth self-limiting.

Self-limiting means additional exposure no longer adds material in an ideal saturated half-cycle. It does not mean gas transport is instantaneous or that every substrate starts growing immediately.

From mechanism to a usable model

For a high-aspect-ratio structure, ask whether reactants reach the bottom and by-products leave. Increasing precursor exposure can improve saturation but costs time; insufficient purging can allow unwanted gas-phase or overlapping reactions.

Compare processes using the actual requirements: allowable temperature, composition, conformality, damage, stress and throughput. A thickness-per-cycle calibration on a flat witness sample does not alone establish complete coverage inside a deep feature.

Thickness≈number of ALD cycles × growth per cycle

Model assumptions: After stable nucleation in a qualified saturation window.

Choose a deposition process from constraints

Define film function and the geometry it must cover.

Eliminate processes exceeding the thermal or contamination budget.

Compare coverage and material quality; validate thickness and composition on representative structures.

Mistake to avoid: ALD is not simply CVD operated at a slow growth rate; the separated, saturating reactions are essential.

Worked problem: A film grows 0.10 nm per cycle after a 10-cycle incubation. How many cycles give 20 nm under this model?

200 productive cycles plus 10 incubation cycles =210. Verify incubation and actual growth by measurement.

Self-test: Why is ALD more conformal than a directional evaporation process?

Self-limiting reactions can cover accessible surfaces, while line-of-sight flux can shadow sidewalls; transport and purge limits still matter.

A process gives 0.08 nm/cycle after nucleation. What thickness do 250 productive cycles give?

20 nm. Total cycles can be higher if incubation occurs, and measured thickness should confirm the assumed growth regime.

Epitaxy and superlattices

Epitaxy grows a crystalline film with an orientation relationship to a crystalline substrate. Homoepitaxy uses the same material; heteroepitaxy uses a different one. Lattice mismatch can strain a thin coherent layer; beyond a thickness-dependent limit, relaxation may introduce dislocations.

A superlattice alternates thin layers, creating repeated potentials and interfaces. Quantum wells confine carriers along one dimension when barriers and thickness permit. Band offsets, strain and interface roughness all affect the result; matching orientation does not mean matching lattice constant.

Build the physical picture

A thin mismatched film can initially adopt the substrate’s in-plane lattice spacing. It stores elastic strain energy, which increases with thickness. Relaxation through dislocations may eventually reduce that energy, at the cost of introducing defects.

In a superlattice, repeated layers create repeated band-edge variations. Whether carriers behave as isolated wells or coupled states depends on barrier height, barrier thickness and interface quality—not merely on alternating composition.

From mechanism to a usable model

A mismatch calculation is a geometric input, not a prediction of critical thickness by itself. Elastic constants, dislocation geometry and growth kinetics also matter. Thermal expansion mismatch can add strain during cooldown even when growth-temperature lattice matching was good.

Draw conduction and valence band edges separately. An electron well need not be a hole well. Compare confinement energy with thermal energy and barrier height before assuming tightly localized carriers.

Mismatch f=(a_film−a_substrate)/a_substrate

Model assumptions: Compare unstrained reference lattice parameters at a defined temperature.

Read a heterostructure

Calculate the unstrained mismatch and identify growth temperature.

Separate coherent strain from relaxed lattice spacing; identify possible defect consequences.

Draw band offsets and estimate confinement with an explicitly stated model.

Mistake to avoid: A lattice-matched interface is not necessarily electronically transparent: band offsets can still exist.

Worked problem: A film lattice constant is 0.565 nm and substrate is 0.543 nm. Find mismatch.

f=0.022/0.543≈0.0405=4.05%. This is a large mismatch; coherence cannot be assumed for arbitrary thickness.

Self-test: What distinguishes epitaxy from simply depositing a crystalline film?

Epitaxy establishes an orientation relationship with the substrate rather than random grain orientations.

How does halving an ideal quantum-well width affect its first confinement energy?

It increases by four because E1∝1/L². A finite barrier may invalidate the infinite-well approximation as the energy approaches the barrier.

Diffusion in solids

Atoms move through thermally activated microscopic jumps. A concentration gradient creates net flux from high toward low concentration even though individual jumps occur in both directions. Fick’s first law relates flux to gradient; conservation plus that law gives Fick’s second law.

The solution depends on initial and boundary conditions. Constant surface concentration differs from a finite deposited dose. Heating increases diffusivity exponentially, so a short high-temperature step may matter more than a long cool step.

Build the physical picture

Microscopic jumps occur in both directions. A concentration gradient gives more possible departures from the crowded side, producing a net flux. The minus sign in Fick’s law records that direction. Conservation then makes the local concentration change depend on the spatial divergence of flux.

A constant-concentration reservoir and a fixed total dose are different experiments. A reservoir can supply new atoms; a closed finite-dose system can only redistribute the atoms already present.

From mechanism to a usable model

For constant D, substituting J=−D∂c/∂x into ∂c/∂t=−∂J/∂x produces the diffusion equation. If D varies in space or with concentration, retain it inside the derivative.

A width proportional to √t is a statistical scale, not a boundary beyond which atoms cannot exist. For multiple thermal steps, diffusion broadening depends on the integrated diffusivity ∫D(T(t))dt. Do not average temperatures before evaluating an exponential diffusivity.

J=−D∂c/∂x; ∂c/∂t=D∂²c/∂x²; D=D₀exp(−E_a/kT)

Model assumptions: Constant D and one-dimensional diffusion, without reactions or concentration-dependent transport.

Solve a diffusion estimate

State whether dose or surface concentration is fixed.

Use consistent length units and evaluate D at each thermal step.

Calculate the characteristic width and check dose conservation for a finite-dose model.

Mistake to avoid: Doubling diffusion time does not double diffusion distance in the constant-D model.

Worked problem: For D=10⁻¹⁶ m²/s and t=100 s, estimate √(2Dt).

√(2×10⁻¹⁶×100)=1.414×10⁻⁷ m=141 nm. This is an RMS scale, not a sharp diffusion front.

Self-test: Why must boundary conditions accompany the diffusion equation?

They define whether material is supplied, removed or conserved, and select different concentration profiles.

How long must diffusion run to double √(2Dt) if D stays fixed?

Four times as long. Squaring the distance ratio gives the required time ratio.

Electromigration

Electromigration is atomic transport influenced by current flow, often described through electron-wind and other driving forces. Atoms accumulate or leave locations, forming hillocks or voids that can change resistance and break a conductor. Grain boundaries and interfaces may dominate paths.

High current density and temperature often shorten lifetime; geometry and mechanical stress also matter. Black’s equation is an empirical lifetime model for a qualified mechanism, not a universal guarantee. Width alone is insufficient if a local neck or contact concentrates current.

Build the physical picture

A conductor can fail without melting. Current-driven atomic transport removes atoms from some regions and accumulates them elsewhere. A void narrows the conducting area, raising current density and heating locally, which can accelerate further damage.

Atoms also respond to chemical potential, temperature and stress gradients. Mechanical confinement can oppose transport. A complete mechanism cannot be inferred from current density alone.

From mechanism to a usable model

Black’s equation describes empirical lifetime scaling for a qualified structure and failure mechanism. Its prefactor and exponents must come from relevant tests. Comparing two temperatures requires kelvin and the same activation energy units as k.

Lifetime distributions matter: a mean lifetime does not guarantee survival of every line. Current crowding around vias, corners and width transitions can make local conditions much harsher than the average J=I/A suggests.

MTTF=A J^(−n) exp(E_a/kT)

Model assumptions: Empirical model with process-specific A, n and activation energy.

Evaluate an interconnect risk

Find the smallest effective cross-section and calculate local current density.

Estimate local temperature, including neighbouring heat sources.

Apply a qualified lifetime model and discuss uncertainty, distributions and mechanism changes.

Mistake to avoid: A wider remote track cannot repair a narrow contact bottleneck.

Worked problem: If n=2 and temperature is unchanged, what does doubling J do to predicted lifetime?

The ratio is 2⁻²=1/4. Real data must establish whether n=2 applies to the conductor and failure mode.

Self-test: Why can a short narrow segment determine the failure location?

It raises local current density and may produce stronger heating or stress gradients.

At fixed current and thickness, width doubles. With n=2 and unchanged temperature, how does predicted lifetime change?

J halves, so the empirical lifetime rises by a factor of four. Real current crowding and thermal changes must still be checked.

Mesoscopic materials

A structure is mesoscopic when dimensions become comparable to microscopic transport lengths while it still contains many atoms. Mean free path measures scattering distance; phase-coherence length measures how far quantum phase remains correlated. These lengths need not be equal.

Ballistic transport means carriers cross with little scattering; coherent transport preserves interference. A material can be diffusive yet phase-coherent over a smaller region. Temperature and disorder change these lengths, so size alone does not specify the transport regime.

Build the physical picture

Transport length scales answer different questions. Mean free path compares the device to momentum-changing scattering. Phase-coherence length compares it to loss of phase information. Thermal and localization lengths can introduce further boundaries.

In a fully transmitted one-dimensional channel, contacts inject available quantum states. Conductance is quantized in an ideal model even without ordinary resistance distributed throughout the channel.

From mechanism to a usable model

The Landauer picture gives G=(2e²/h)ΣTn for spin-degenerate channels, where Tn is each transmission probability. Channels add conductance in parallel. Imperfect contacts reduce transmission; opening two channels is not equivalent to placing two resistors in series.

Temperature broadens the energy range occupied near the Fermi level. Disorder can make transport diffusive while coherence remains over part of the sample. Always state which length is compared with which physical scale.

G₀=2e²/h≈77.5 µS

Model assumptions: Ideal spin-degenerate fully transmitted channel; real contacts and transmission probabilities modify conductance.

Classify a small conductor

Compare active length with mean free path and coherence length.

Identify the number of available channels and their transmission assumptions.

Predict conductance and distinguish contact effects from internal scattering.

Mistake to avoid: Nanoscale is a size label, not proof of quantum-coherent or ballistic behaviour.

Worked problem: Find resistance of one ideal conductance quantum.

R=1/G₀≈12.9 kΩ. Several fully transmitted channels add conductance, not resistance.

Self-test: Can a nanowire always be called ballistic?

No; compare its active length with the mean free path under the actual conditions.

Two spin-degenerate channels have transmissions 1 and 0.5. Estimate G and R.

G=1.5G0≈116.2 µS; R≈8.60 kΩ. This is the ideal two-terminal Landauer result under the stated channel model.

Carbon nanomaterials

Carbon’s bonding permits very different structures. Diamond has a three-dimensional sp³ network; graphite consists of sp² layers. Graphene is a single such layer, nanotubes are rolled structures, and fullerenes are closed cages. Geometry and electronic structure change properties without changing the element.

Nanotube chirality influences electronic behaviour; pristine graphene lacks the useful conventional band gap of a silicon switch. Contacts, defects, substrate interactions and functionalization can dominate a fabricated device. High intrinsic mobility does not by itself guarantee a manufacturable transistor.

Build the physical picture

Carbon illustrates why composition alone is insufficient. Changing hybridization and connectivity changes available electronic states. Graphite’s layered bonding also makes many properties directional; a single isotropic number can conceal the important behaviour.

A nanotube’s chiral vector determines how the graphene lattice is wrapped. Quantization around its circumference selects allowed states, helping explain why different tubes can have different electronic character.

From mechanism to a usable model

Intrinsic mobility is only one device metric. A switch also needs a useful off-state, reliable contacts, controlled channel dimensions and compatible fabrication. Defects can scatter carriers but can also enable chemical sensitivity.

Raman features, microscopy and electrical measurements probe different aspects of a carbon sample. No single peak or micrograph establishes all of purity, layer count, defect density and contact quality. Interpret characterization against the actual application.

Nanotube circumference=|n a₁+m a₂|

Model assumptions: Chiral-vector description for an ideal graphene lattice.

Compare carbon candidates

Specify the bonding and geometry rather than only the element.

Connect electronic states to the desired conduction or switching function.

Account for defects, contacts and fabrication before comparing device performance.

Mistake to avoid: High carrier mobility does not imply high digital on/off ratio.

Worked problem: Why can two carbon samples have opposite electrical behaviour?

Bonding and structure differ: diamond’s wide gap contrasts with graphite’s electronic states. Elemental composition alone does not determine transport.

Self-test: Why is graphene’s lack of a conventional gap a switching challenge?

It makes strong current suppression difficult in a simple field-effect device.

Why can a graphene chemical sensor benefit from surface exposure that harms a transistor’s reproducibility?

Adsorbed molecules can change carrier density and conductance, providing sensitivity. The same uncontrolled adsorption creates drift and variability when stable switching is the objective.

Liquid crystals

Liquid crystals combine molecular order with fluid motion. Nematic phases have preferred orientation; smectic phases also have layering; cholesteric phases exhibit a twisted orientational pattern. The director describes average orientation, not a permanent arrow attached to each molecule.

An electric field can rotate the director when dielectric anisotropy permits, changing optical response. Alignment layers and elastic forces compete with the field. Displays convert this reorientation into intensity through polarizers and cell geometry; voltage is not emitted light.

Build the physical picture

The nematic director represents an average molecular axis; reversing it gives the same orientation state for an ordinary nematic. Elastic energy penalizes splay, twist and bend, while surface anchoring favours particular boundary orientations.

An applied field competes with those elastic and anchoring energies. Dielectric anisotropy determines whether molecules tend to align parallel or perpendicular to the field.

From mechanism to a usable model

Optical readout depends on birefringence and polarization. Reorientation changes the phase accumulated by polarization components; polarizers turn that phase change into intensity. A liquid-crystal cell does not ordinarily generate the light itself.

The approximate relaxation time γ1d²/K follows from balancing viscous resistance against elastic restoration. Field-driven turn-on can have a different timescale. Temperature influences viscosity, order and phase, so thickness scaling alone is not a full response model.

Characteristic response time τ∝γ₁ d²/K

Model assumptions: Simple elastic-viscous scaling with rotational viscosity γ₁, cell thickness d and elastic constant K.

Explain a display transition

Identify initial anchoring, director orientation and polarization.

Describe the field-induced orientation and resulting optical phase change.

Separate electrical driving time from mechanical reorientation and optical readout.

Mistake to avoid: A director is not an electric dipole arrow, and all liquid-crystal modes do not respond identically.

Worked problem: If cell thickness doubles with material properties fixed, how does this timescale change?

It rises by a factor of 4. Actual turn-on dynamics also depend on voltage and boundary conditions.

Self-test: Why does an LCD need an optical arrangement as well as electrodes?

Reorientation changes polarization or birefringence; polarizers and illumination convert that change to visible intensity.

If viscosity doubles while thickness and elasticity remain fixed, what happens to the simple relaxation time?

It doubles. This applies to the stated viscous-elastic relaxation model, not automatically to every driven transition.

Functionally graded materials

A graded material changes composition or microstructure gradually across space. This can smooth thermal expansion, stiffness or electronic band-edge transitions compared with an abrupt interface. The gradient is itself a design variable requiring controlled processing and characterization.

Grading does not eliminate all residual stress or interdiffusion. A useful model must specify how local properties depend on composition and whether each slice remains compatible with adjacent layers.

Build the physical picture

A graded layer replaces a sharp property jump with a spatial profile. This can distribute strain, alter band-edge forces or tailor heat transport. It does not make the total mismatch vanish; constraints still determine how the structure can deform.

For a composition-dependent property, define the relation between composition and local response. Linear interpolation may be an approximation, and microstructure can change that relation.

From mechanism to a usable model

Through-thickness heat passes through successive slices, so their resistances add. Along-plane heat can instead see parallel paths. This geometric difference explains harmonic versus arithmetic averages.

For thermal stress, freely expanding layers and bonded constrained layers behave differently. A graded CTE profile requires a mechanical compatibility analysis, not simply a comparison of endpoint coefficients.

R_th=∫dx/[k(x)A]

Model assumptions: One-dimensional steady conduction, constant area and negligible contact resistance.

Build a graded-medium model

Define x, composition c(x), and local property k(c).

Choose series or parallel geometry and discretize into thin slices if necessary.

Integrate the response and check limiting cases against uniform materials.

Mistake to avoid: A smooth composition profile does not guarantee smooth stress or defect-free interfaces.

Worked problem: Two equal-thickness layers have thermal conductivities 10 and 100 W/(m·K). Is the through-thickness effective value 55?

No. Series thermal resistance gives the harmonic average: k_eff=2/(1/10+1/100)=18.18 W/(m·K). An arithmetic average fits a different parallel geometry.

Self-test: What advantage can gradual CTE variation provide?

It can distribute mismatch over a region, although constraint and total expansion still control stress.

A uniform 1 mm layer with k=10 W/(m·K) and area 1 cm² carries 1 W. Find its temperature drop.

Rth=L/(kA)=0.001/(10×10⁻⁴)=1 K/W, so ΔT=1 K. Interface resistances would add to this result.

Material compatibility during fabrication

A process must remove or grow one material without damaging what must remain. Etch selectivity compares removal rates; masks need enough thickness to survive the required etch. Adhesion, stress, chemical resistance and thermal budget create additional constraints.

Process order matters because later heating may diffuse dopants or react contacts. Plasma exposure can alter interfaces. A successful single-layer process may fail on a multilayer stack, so evaluate the actual interfaces and sequence.

Build the physical picture

Fabrication is a sequence, not a collection of isolated successful processes. A film must survive later heating, chemicals and plasma exposure. Mask adhesion and pinholes can matter as much as nominal selectivity.

Etch rate can depend on feature density and aspect ratio. A blanket sample rate may not represent a narrow trench or a loaded wafer. End-point detection and overetch compensate for variations but introduce their own costs.

From mechanism to a usable model

Calculate mask consumption using the longest required etch time, including overetch. Then allow a margin for thickness variation, roughness and local erosion. Selectivity alone does not describe sidewall shape or damage.

Build a compatibility matrix for target, mask and underlying layers. Include thermal diffusion, unwanted reactions, contamination and mechanical stress. Process order should protect the most vulnerable interfaces.

Selectivity S=etch rate of target / etch rate of mask

Model assumptions: Rates depend on chemistry, temperature, geometry and loading.

Check a process window

Determine clearance time for the slowest region.

Calculate target, mask and underlying-layer loss over that time.

Compare all losses and temperatures with allowed budgets; qualify on the real stack.

Mistake to avoid: A chemically selective etch can still damage an electronic interface through charging or ion bombardment.

Worked problem: Etch 500 nm of target at 100 nm/min with a mask etched at 5 nm/min. Find minimum consumed mask thickness.

Time=5 min; mask loss=25 nm. Additional thickness is needed for nonuniformity and overetch. Selectivity=20.

Self-test: Why add an overetch, and what does it cost?

It clears slow regions but consumes extra mask or underlying material and can worsen profile damage.

The 500 nm etch takes 5 minutes plus 20% overetch. Mask rate is 5 nm/min. How much mask is consumed?

Six minutes consume 30 nm. The starting mask must exceed this with a qualified margin.

Functional materials

A functional material converts or controls a physical signal. Piezoelectric response couples strain and electric polarization; ferroelectrics possess switchable spontaneous polarization. Magnetic order and optical transitions create other sensing and actuation mechanisms.

A ferroelectric can be piezoelectric in an appropriate state, but the words are not interchangeable. Hysteresis, domains, temperature and fatigue affect operation. Use constitutive coefficients with the correct direction, frequency and boundary conditions.

Build the physical picture

Constitutive equations connect causes and responses, but boundary conditions choose which coefficient you measure. A mechanically clamped piezoelectric element behaves differently from a free one. Tensor direction also matters: a coefficient along one axis cannot automatically be used along another.

Ferroelectric polarization can remain after the external field is removed. Domain switching creates a history-dependent loop, unlike a purely linear dielectric response.

From mechanism to a usable model

For a simple free thickness actuator, strain=dE and E=V/L combine to give ΔL=dV. The cancellation of L belongs to this ideal geometry; it does not mean dimensions are irrelevant for force, capacitance or breakdown.

Sensors and actuators trade displacement, force, bandwidth and loss. Hysteresis may support memory but complicate precision positioning. Report frequency, temperature, preload and field amplitude when comparing coefficients.

D=d·stress+εE; strain=s·stress+dE

Model assumptions: Linearized piezoelectric relations; tensor directions and hysteresis may matter.

Choose a functional-material model

Identify electrical and mechanical boundary conditions and tensor direction.

Use a small-signal relation only within its appropriate range.

Add domain, hysteresis or resonance effects when the operating conditions demand them.

Mistake to avoid: Piezoelectricity and ferroelectricity are related but not synonyms.

Worked problem: For an ideal actuator coefficient d=200 pm/V and voltage 10 V across its active direction, estimate displacement in the simple thickness mode.

ΔL=dV=2 nm. Real stacks amplify displacement through multiple layers and can be mechanically constrained.

Self-test: What makes a ferroelectric hysteresis loop useful and problematic?

It supports memory through remanent states, while switching loss, fatigue and history complicate analogue response.

For d=100 pm/V, estimate free thickness displacement at 20 V.

ΔL=dV=2000 pm=2 nm. This does not specify blocked force or motion of a constrained device.

Energy materials

Energy devices depend on transport and chemical or electronic free-energy changes. Photovoltaics separate photogenerated carriers; batteries move ions internally and electrons through an external circuit. Catalysts change reaction pathways and rates without changing equilibrium free-energy differences.

Performance depends on interfaces, diffusion distances and stability as well as active material. Degradation can consume electrolyte, trap carriers, crack particles or grow resistive layers. Compare usable energy, power and retention under defined conditions.

Build the physical picture

Separate thermodynamic capacity from transport-limited rate. A battery can contain a large amount of active material but deliver little of it quickly if ions cannot reach reaction sites. Electrons and ions travel through different pathways that must both remain connected.

A catalyst reduces a kinetic barrier; it does not supply a favourable equilibrium free-energy change. Similarly, a photovoltaic needs absorption, charge separation and collection—not just a material that absorbs light.

From mechanism to a usable model

Specific energy divides usable energy by total device mass, including inactive components. Capacity alone misses operating voltage; peak power alone misses duration. Measure under a stated temperature, rate and cutoff condition.

Nanosizing shortens diffusion times approximately as L²/D, but increases exposed surface and therefore potential side reactions. A material improvement must be evaluated against both rate and long-term stability.

Energy=∫V I dt; ideal battery capacity=nF/3600 in Ah per mole of reaction

Model assumptions: n is transferred electrons per reaction unit; practical capacity includes inactive material and incomplete utilization.

Compare two energy materials

Define the device-level metric and measurement conditions.

Identify limiting charge transport, reaction and collection mechanisms.

Account for inactive mass and degradation before ranking performance.

Mistake to avoid: More surface area is not unconditionally better; it can increase unwanted reactions as well as useful ones.

Worked problem: A 3.7 V cell supplies 2 A for 30 minutes at roughly constant voltage. Estimate energy.

Charge=1 Ah; energy≈3.7 Wh=13.32 kJ. Voltage variation and losses require integrating the real curve.

Self-test: Why can nanosizing improve rate while worsening stability?

Shorter transport distances and larger area help kinetics, but more exposed interface can accelerate side reactions.

Halving a diffusion distance leaves D unchanged. What happens to the characteristic diffusion time?

It falls by four because t scales as L²/D. Other rate limits may still dominate the complete device.

More in Micro & Nano

• Semiconductor Devices 1Derive carrier populations and transport, then connect electrostatics to diodes, BJT and MOS devices.
• Electronics Technology 1How requirements become a manufactured, measurable and reliable electronic assembly.
• Semiconductor FabSemiconductor fabrication and cleanroom technology: silicon wafers, RCA cleaning, oxidation, lithography, etching, doping, PVD/CVD/ALD, metallization, CMP, packaging, test and the complete CMOS flow from sand to transistor.
• NanodiagnosticsFollow excitation through sample interaction and detection to a defensible material conclusion.
• NanotechnologyWhy size changes behaviour, and how nanostructures are patterned, grown and integrated.
• Microscopy Start HereA beginner-first path through optical microscopy, SEM, TEM, AFM and STM with acronym decoding, instrument anatomy, physics, workflows and artifact tests.
• GlossaryProfessional characterization terminology, units and distinctions.
• Sample PreparationPreparation choices that alter surfaces, interfaces and electrical results.
• Core FoundationsA guided foundation path through mathematics, electronics technology, semiconductor devices, materials, nanotechnology and scientific practice.
• Units, Fields & EnergySI units, electron-volts, gradients, conservation, exponentials and uncertainty with worked calculations.
• Atoms, Crystals & BandsBonding, crystal planes, defects, reciprocal space, band formation and effective mass.
• Contacts, MOS & TransistorsWork functions, Schottky and ohmic contacts, MOS electrostatics, threshold and transistor operation.